REVIEW

From Movement to Thought: Executive Function,Embodied Cognition, and the Cerebellum

Leonard F. Koziol & Deborah Ely Budding &

Dana Chidekel

# Springer Science+Business Media, LLC 2011

Abstract This paper posits that the brain evolved for thecontrol of action rather than for the development ofcognition per se. We note that the terms commonly usedto describe brain–behavior relationships define, and inmany ways limit, how we conceptualize and investigatethem and may therefore constrain the questions we ask andthe utility of the “answers” we generate. Many constructsare so nonspecific and over-inclusive as to be scientificallymeaningless. “Executive function” is one such term incommon usage. As the construct is increasingly focal inneuroscience research, defining it clearly is critical. Wepropose a definition that places executive function within amodel of continuous sensorimotor interaction with theenvironment. We posit that control of behavior is theessence of “executive function,” and we explore theevolutionary advantage conferred by being able to antici-pate and control behavior with both implicit and explicitmechanisms. We focus on the cerebellum's critical role inthese control processes. We then hypothesize about theways in which procedural (skill) learning contributes to theacquisition of declarative (semantic) knowledge. Wehypothesize how these systems might interact in the processof grounding knowledge in sensorimotor anticipation,thereby directly linking movement to thought and “embod-ied cognition.” We close with a discussion of ways inwhich the cerebellum instructs frontal systems how to think

ahead by providing anticipatory control mechanisms, andwe briefly review this model's potential applications.

Keywords Cerebellum . Executive function . Embodiedcognition . Frontal systems . Forward models . Inversemodels

Introduction

Studies considering the contributions of various brainregions to adaptive behavior frequently conceptualize thesystems governing brain–behavior relationships withinseparate and distinct domains. Clinical neuroscience, forexample, often views behaviors as differentiated betweencognitive, attention/executive, language, visuospatial, learn-ing and memory, and sensory and motor domains.However, these distinctions are artificial, and they obscurethe fact that the brain operates as an integrated whole.Moreover, the longer we compartmentalize brain functionin this modular way, the more habitual this way of thinkingbecomes. The constructs reify and we may not realize howmuch they constrain our investigations and our ability toapply research findings to advance understanding of brain–behavior relationships [1]. This is not only of theoreticalimportance, but it is also of practical significance because italso limits our approaches to treating patients.

The prevailing models that guide understanding ofbehavior and cognition within certain clinical fields suchas neuropsychology, psychiatry, neurology, and occupation-al and physical therapy are “top down” or cortico-centric innature. In emphasizing cortical function, these fields fail toconsider the substantial roles that subcortical structuresplay, which other branches of neuroscience have increas-ingly recognized as critical to adaptation [2–4].

L. F. Koziol (*)Chicago, IL, USAe-mail: lfkoziol@aol.com

D. E. BuddingManhattan Beach, CA, USA

D. ChidekelTarzana, CA, USA

CerebellumDOI 10.1007/s12311-011-0321-y

The INS Dictionary of Neuropsychology defines “cog-nition” as, “the mental processes associated with attention,perception, thinking, learning, and memory” [5]. Thisdefinition is so all-encompassing as to render it practicallymeaningless. Of what practical value is a term such as“cognition” if it must be operationally defined in each andevery instance it is applied [1]? Such a broad definition of“cognition” encompasses the concept of executive function,while at the same time, the term “executive function” (EF)is defined in different ways by the clinical fields that applyit. Even neuropsychiatry and neuropsychology are unableto agree upon the key elements that comprise this obviouslymultiple component function (see definitions provided byLezak, Gualtieri, and Dubois et al. [6–8]). We are definingEF as the functions an organism employs to act indepen-dently in its own best interest as a whole, at any point intime, for the purpose of survival, as initially proposed byMiller [9].

While there are no universally agreed upon definitionsfor “cognition” or “executive function,” both have histor-ically been conceptualized as falling under the directinfluence of conscious voluntary control. More recently,efforts have been made to include emotional and motiva-tional functions within conceptualizations of EF [10, 11].The prefrontal cortex (PFC) has been described asparticipating in two separate but related categories of EF.“Metacognitive” EFs include problem solving, planning,concept formation, strategy development and implementa-tion, and controlling attention and working memory; thesefunctions depend upon the integrity of the dorsolateralprefrontal cortex (DLPFC). “Emotional/motivational” EFsinclude coordinating cognition with motivation and affectin order to fulfill biological needs in relation to theconditions at the time; these functions depend upon medialfrontal and orbitofrontal areas. They are believed to bephylogenetically older and perhaps more primitive EFs, butthey nevertheless remain essential to successful, autono-mous adaptation [11]. These concepts are very similar tothe concepts of “cold” and “hot” EF, which are dependentupon the functions of the cognitively based DLPFC and themore medial PFC cortices, respectively [12].

The neuroanatomical systems that are cited in support ofthese proposals are consistent with conceptualizations ofthe limbic system as consisting of a phylogenetically olderorbitofrontal division and a newer hippocampal division[13]. The phylogenetically older division makes executivedecisions that are intuitive, based upon processes and“hunches” of which we are not consciously aware.Instrumental learning, in which the basal ganglia play suchan important role, is clearly implicated in this system [14,15]. The newer hippocampal division presumably providesthe context in which decisions and plans are made on thebasis of conscious control. This is the substrate for goal-

directed thinking, which is considered to be under theinfluence of conscious awareness. However, EF is not onlyassociated with thinking but with “visceral” functions aswell. This is a critical point. Cortico-centric models ofbehavior seem to assume that “intention” and “goal-directed” behaviors are primarily, if not solely, the productsof deliberation and conscious thought that are mediatedexclusively by the neocortex and specifically by theprefrontal lobes. However, as models of behavior evolvethat integrate the role of subcortical structures in theseprocesses, our understanding of “cognition” and “EF” isexpanding to include the ways in which aspects of thesefunctions operate outside of conscious awareness.

Serial-Order Processing, Conscious Cognitive Control,Executive Functions, and Continuous SensorimotorInteraction

Action (or movement) and goal-directed behavior areinherent in the concepts and definitions of cognition andEF. Conscious cognitive control of action and goal-directedbehavior have been emphasized in these conceptualizations,while functions that underlie human conscious and non-conscious processing have not been included. Wereconscious control necessary for executive function, onewould imagine that executive function would be unique tohumans, but it is not. Other animals clearly demonstrate“EF” in their behavior. Non-human primates use tools. Thismeans they plan, which in turn means they have some formof “EF”[16, 17]. In fact, Ramnani [18] has recentlyreviewed the prefrontal-cerebellar and posterior parietal-cerebellar connectional profiles in non-human primates,which appear to represent the anatomic underpinning for“EF.” This neuroanatomy includes many aspects of thecircuitry that we propose supports EF, embodied cognition,and the influence of cerebellar control models in humansensorimotor learning. Additionally, other animals withoutlanguage capability can imagine and compare differentnavigational pathways before embarking on them in practice,and they can select from certain short-term and long-termgoals, while they also predict and avoid certain dangers [19].

Such findings portray that EF cannot possibly be uniqueto human conscious cognitive control and cannot be solelydependent upon declarative knowledge and languageprocesses. So how do animals engage in these behavioralselections that are consistent with conceptualizations of EFas applied to humans? Pezzulo [20] has attempted toaddress this question theoretically by proposing that allknowledge for behavior, for all organisms, is derived fromsensorimotor anticipation. His conceptualization is basedupon three ideas. First, any organism's knowledge andrepresentational ability is derived from sensorimotor inter-

Cerebellum

action. This idea is similar to early stages of Piaget'sdevelopmental theory. In psychological, non-neuroscienceliterature, it is often referred to as “embodied cognition”[21, 22]. Second, he asserts that the reuse of sensorimotorlearning mechanisms throughout the course of evolution issufficient to account for the development of procedural anddeclarative knowledge. In this model, the anticipatorymechanisms that are used in the control and execution ofbehavior generate procedural knowledge. At the same time,the re-enactment of these situated behaviors and actionsthrough mental simulation or imagination can producedeclarative knowledge. Third, these functions or processesallow for internal manipulation that can generate “EF.” Thisis the result of imagining the results of prior activities,which he describes as internalized situated action.

Pezzulo's theoretical position is compelling, but it is notsystematically superimposed upon a functional neuroanat-omy. This paper will consider the neuroanatomic under-pinnings of the ways in which sensorimotor interactionsupports procedural and semantic, declarative knowledge.This viewpoint will provide biologic consistency along thephylogenetic scale in an attempt to understand “cognition”as action control or “EF.” The current paper will focus onthe specific role of the cerebellum at the expense ofcritically examining the important roles played by otherbrain regions, which have been described in great detailelsewhere [23–26].

It has been estimated that much of what we do, perhaps95% of an adult human's activity or behavior is routine andautomatic, outside of voluntary conscious cognitive aware-ness [27–29]. These are things we do spontaneously,“without thinking,” simply because they need to be done.These behaviors are not merely due to faster processing ascompared to controlled processing [30]. Instead, theseautomatic behaviors are efficient, economical, elegant,initiated without uncertainty or hesitation, and alwaysadaptive in context and adjustable to contextual changes[31]. Such behaviors can be construed as executivefunctions. Their efficient, automatic quality demonstratesthe significant role implicit processes play in EF. This pointis critical to understanding our view of EF, its multifacetednature, and its anatomic underpinnings.

Such automatic behaviors cannot possibly be accountedfor by a model of cognitive control that depends uponserial-order processing. Within this traditional (and stillpopular and clinically applied) model, three processes arecentral: first, we perceive; next, we “think” to formulate orchoose a response; last, we act. This processing modelrelies upon direct sensorimotor feedback, which is inevita-bly delayed by numerous factors, including the timenecessary for processing sensory information, and the delayfor transmission of motor commands from the brain to themuscles [32, 33]. While some behaviors are generated this

way, this means of processing is too time and resource-consuming to account for how smoothly and quicklypeople are generally able to coordinate and control theirbodies and objects within the external world. Seminalpapers by Cisek and Kalaska, Shadlen and Movshon, andSinger review neurobiologic data from various disciplinesand conclude there is little evidence to support aperception-cognition-action model as a phylogeneticallyconserved and primary mode of adaptation [34–36].

This consciously driven, top down model's insufficiencyfor explaining so many of our adaptive behaviors points tothe need for an alternate model. A more implicit, bottom-upmodel that emphasizes the role of subcortical brainstructures (e.g., the “bottom” in “bottom up”) and is basedupon continuous sensorimotor interaction with the environ-ment fills this need. Such a model provides phylogeneticand biologic continuity across species, in most of whom thebrain clearly did not evolve for the primary purpose ofcognition or “thinking” [37]. Instead, the brain evolved tomeet changing and emerging demands for survival, in thecontext of continuous interaction with the environment.Therefore, there must have been considerable evolutionarypressure to develop the ability for anticipation to guide thephysical actions necessary for survival. Evolutionaryprocesses favored the development of predictive or “antic-ipatory,” and simulative or “imaginative” mechanisms forthe purpose of action control and not for cognition per se.Both procedural (skill) knowledge and declarative (seman-tic) knowledge can thus be explained and understood interms of continuous “on-line” sensorimotor anticipation and“off-line” simulation of (imagination or thinking about)potential action. Animals that possessed only mechanismsthat allowed them to adjust to behavior as circumstancesunfolded could not survive and adapt because theirbehavior would be limited to reactive context processingin the moment, in the here-and now. To survive, they wouldalso have to develop mechanisms for proactive contextprocessing to serve intentional control of goal-directedbehavior. These processes lay at the base and heart of EF.

Bloedel and Bracha posit that there is no duality betweenmotor and “cognitive” functions [38], and we agree.Without movement or action, there is no need for thought.We (and numerous others) argue that for the brain, theprimary substantive difference between planning an activityand engaging in its motor counterpart is the actualexecution of that behavior [4, 39–41]. For example, thereis no inherent difference between planning how torearrange the furniture in one's living room and actuallydoing it. Thought manipulation controls action. Thoughtevolved from the motor system as a mechanism to facilitatethe development of motor programming and action control.

In the following sections of this paper, we emphasizethat the cerebellum plays a critical role in sensorimotor

Cerebellum

behavior. As such, it lays at the foundation of executivefunction as defined as an organism's acting independentlyin its own best interest as a whole, for the purpose ofsurvival [9]. We argue that the cerebellum appears to trainfrontal systems about how to execute motor behavior invarious settings and contexts by anticipating outcomes orsensory/knowledge feedback. Therefore, significant frontallobe “knowledge” is situated in a sensorimotor format. Wepromote an active rather than passive position for thecerebellum due to the role the cerebellum plays ininstructing or teaching the frontal lobes how to predict(anticipate) by “thinking” about movement outcomes [42].

A phylogenetic model portrays new structures andfunctions emerging in the course of evolution by buildingonto existing structures [43]. As the cerebellum developedin animals long before frontal cortex expanded, a role forthe cerebellum in the function of frontal systems is inherent.Frontal systems have historically been considered those thatsupport the ability to plan and think ahead, and in this wayare considered to support abstract thought. We argue thatthese higher-order abstract thinking skills comprise anextension of the motor control system that is situated inbasic anticipatory control mechanisms. We posit thatcerebro-cerebellar functions play a critical role in allowingfrontal systems to plan, to think ahead, and to abstract. Inorder to lay the foundation for this position, we firstsummarize a model of adaptation based upon continualsensorimotor interaction with the environment.

Sensorimotor Interaction with the Environment

The vertebrate brain's functional architecture evolved to meetthe needs of interactive behavior. This evolutionary trendwas strongly conserved throughout the course of phylogeny.As summarized by Cisek and Kalaska and others, ourperceptions of the world are not the result of a serialprocessing or reconstruction processes that use sensoryinformation to build or develop an internal representationof the external world. Instead, neural processing is contin-uous, while the salient properties of objects, such as whatthey look like and what they are used for, are represented inthe same sensory and motor brain circuits that were activatedwhen the information about that object was initially acquired[44–48]. These neuroanatomic “facts” can be referred to as a“cerebral cortical model,” consisting of body and motorschemas as described by Ito, or praxicons and innervatoryprograms as described by Heilman and Rothi [49, 50].Cognitive psychologists might refer to this as “embodiedcognition.” Within this information processing model, theventral “what” pathway and the dorsal “where” pathwayhave additional functions that go well beyond the realm ofobject recognition and object location, respectively.

Processing within the parietal cortex dorsal stream andreciprocally connected premotor regions is primarilyconcerned with pragmatic, practical representations of theopportunities for action that those objects afford or offer[51–53]. This dorsal pathway registers not only wheresomething is, but also “how to do” something. How tograsp, for example, might be determined by an object'sshape, size, and movement. This region plays a critical rolein procedural memory for action concepts. The posteriorparietal cortex (both superior and inferior regions) and itsreciprocal connections with premotor areas and the frontaleye fields transform the dorsal pathway into a visual controlarea for actions. The parietal cortex focuses upon spatialinformation because this data is critical for specifying theparameters of ongoing and potential actions. It operateswithin egocentric or “reaching” space [54]. (We will returnto these critically important points about the dorsalinformation processing stream in subsequent sections.)

The ventral information processing stream registers“what” the individual is seeing and information about“what” that object is used for, which is essentially asemantic function that can operate in allocentric space. Thispathway operates relatively slowly in contrast to thequickly responding dorsal pathway, since the ventralpathway does not primarily serve an “action” purpose[54]. However, because it specifies the reward value ofobjects (to be described below) it plays an important role inbehavioral praxis. The dorsal and ventral pathways havecallosal connections and project to the hippocampalformation for memory storage. This overlap allows thesystems to operate in a complementary fashion. These twosystems are believed to be integrated by the age of9 months, which has obvious relevance for theories ofpediatric neurodevelopment [55].

The ventral pathway provides information for action orbehavioral selection by biasing potential actions withinformation about reward value associated with objectidentity. This behavioral biasing (which is essentially aform of anticipating or predicting the outcome of abehavior) includes information from basal ganglia rewardcenters and regions of the prefrontal cortex that predictreward outcomes [56–59]. Several potential actions areavailable in most situations. Accordingly, these potentialactivities, choices or decisions are reflected over largeportions of the cerebral cortex. Decision making is thus notstrictly localized within the prefrontal cortex. Instead, it isfound within the same sensorimotor circuits that areresponsible for processing sensory information, associatinginformation with reward value, and programming andexecuting the associated actions. This organizational profileallows individuals to engage in a level of adaptivefunctioning that is characterized by automatic behaviorsalternating with episodes of higher-order control. In an

Cerebellum

unpredictable environment, the adaptive value of suchflexibility cannot be overemphasized.

Within this model, “cognition” is not separate fromsensorimotor control [34, 60]. In fact, sensorimotor controlis primary and paramount, while conscious cognitivecontrol becomes subordinate. The final “decision,” or theselected action/behavior, is an outgrowth of cortical-basalganglia interactions. While behaviors result in overtfeedback from the environment, action is undertaken inconcert with anticipated or predicted feedback through thecerebellum, which appropriately adjusts the behaviorthrough context-response linkage [61]. This model empha-sizes sensorimotor interaction and the brain's inherentcapacities to predict and anticipate. This allows behavior tooccur in “real time” in a manner that is not easily explainedthrough a static “perception-cognition-action” model. AsKinsbourne and Jordan note, anticipation is continuous and

ongoing; the ability to anticipate is an inherent designcharacteristic of the brain [62]. This sensorimotor paradigmis illustrated in Fig. 1 (For a comprehensive review of thismodel, see Cisek and Kalaska [34]). The remainder of ourdiscussion focuses upon the specific role of the cerebellumwithin this model of adaptation.

The Cerebellar Anatomic Underpinningfor Sensorimotor Interaction and EF

Cerebellar contributions to motor and non-motor behaviorare made possible through the cerebro-cerebellar circuitrysystem. The importance of this system can be inferred bytracking its evolutionary development. Smaers, Steele, andZilles examined the cortico-cerebellar system in 19 anthro-poid species spanning across 35 million years of divergent

Fig. 1 This diagram represents a sensorimotor interaction paradigmfor visually guided movement, illustrating interactions between thecortex, basal ganglia, and cerebellum. The dorsal stream, whichconsists of the parietal cortex and reciprocal connections withpremotor regions, is concerned with practical representations orprograms for a situation's action opportunities. The dark blue arrowsrepresent action specification, which comprises the parameters forperforming the behavior. This processing originates in the occipitallobes and proceeds in a rightward direction through the dorsalpathway. This dorsal pathway registers not only where something is,but also “how to do” something—such as how to grasp—as might bedetermined by an object's shape, size, and movement. This regionplays a critical role in the procedural memory for action concepts. Theventral stream serves object identity (connections between the ventral

and dorsal stream are not directly illustrated; see text). The polygonsrepresent three potential actions (neural populations) along the dorsalroute. Competition for action choice or selection is biased by “rewardcenter” input from the basal ganglia to parietal, temporal, andprefrontal neocortical regions. Reciprocal red arrows represent thisevaluative biasing process. Therefore, action choices are representedover large regions of cerebral cortex and subcortical regions. Theaction that is selected depends upon the behavior with the strongestbias. The cerebellum (blue and dotted blue arrows connecting thecerebellum with the cortex in this drawing) adjusts behavior on thebasis of internal anticipated or predictive feedback that are inherent incerebellar control models, which are not illustrated in this diagram andare the subject of the remainder of this paper

Cerebellum

evolution. They concluded that neural systems involvingprofuse cortico-cerebellar connections composed a majorfactor in the evolution of anthropoid brain organization[63]. In humans, nearly every region of the cerebral cortex(with the exception of the inferotemporal cortex) hasreciprocal, segregated connections with the cerebellum(for a review, see Schmahmann and Pandya [64]). Whilethe neocortex has expanded during the course of phylogeny,the cerebellum has also demonstrated a three- to fourfoldexpansion in sapiens compared to other species [65]. Alongwith the selective expansion of the cerebellar cortex [66],the dentate nucleus of the cerebellum has dramaticallyincreased in size, and reciprocal connections between theprefrontal cortex and the dentate nucleus have beenidentified in humans that are not present in other species[67–69]. There is compelling neuroanatomic evidence for acerebellar role in a variety of processes, including attention,executive functioning, learning and memory, visuo-spatialregulation, language, and affective-behavioral modulation[70–72]. Balsters and Ramnani [73] have also providedsupportive evidence for the view that prefrontal projectingareas of the cerebellar cortex process information that is ofa purely abstract nature.

While the reciprocal, segregated, closed loop circuitryprofile represents the major architectural unit of cerebro-cerebellar interactions, the uniformity of the intrinsiccellular organization of the cerebellum implies uniformityin information processing by the cerebellum, regardless ofthe area in the cerebral cortex from which the information isreceived [74]. In other words, the cerebellum performs thesame operation on information regardless of its source. Thisis consistent with Ito's well-recognized proposal that for thecerebellum, movement and thought are equivalent; thecontrol and manipulation of thought content is no differentfrom the control and manipulation of body limbs inproblem solving [4]. Once a movement, perception, orthought is “coded” within the neural circuitry of thecerebellum, the input is manipulated in an identical way.The simplistic idea that equates advancement of cognitiveskill with neocortical expansion must be revisited inconsideration of the neocortex's and cerebellum's coordi-nated expansion as a functional ensemble or network [75].

Cerebellar Control Models

As referenced above in a previous section, direct corticalsensory feedback processes operate too slowly for us togenerate effective behaviors in “real time” [33, 76].Therefore, brain mechanisms have evolved that allow theorganism to predict or anticipate sensorimotor feedback inorder to generate rapid, automatic, adaptive behaviors.Within the cerebellum, these mechanisms are referred to as

internal models [77]. The cerebellum constructs two formsof internal models, namely, a forward model and an inversemodel. Their initial development is made possible throughthe “mossy fiber” cerebro-cerebellar circuitry system,which essentially allows the cerebellum to copy thecontents of cortical, conscious working memory and in sodoing, allows the cerebellum to “know” what the cerebralcortex wants to do.

For an adult learning a new task, the first step of learningproceeds consciously and involves neural signals from thepremotor and primary motor cortices and the temporo-parietal cortex (Ito, page 184) [49]. Initial learning trialsrely on external sensory feedback. As learning proceeds,the parietal cortex acquires a “body schema” and a “motorschema” necessary for the task. In Heilman's terminology,these schema are referred to as praxicons and innervatoryprograms, respectively [50]. The cerebellum copies theseschemata in the formation of forward and inverse models.

Therefore, a cerebellar internal model consists of all ofthe dynamic sensory and motor processes necessary toperform a movement or behavior. This internal model isformed and adjusted as a movement is repeated. Behavioralrefinement occurs because the cerebellar olivary systemdetects errors or imprecision in performance and allows forcorrection in predicted feedback and adjustment in move-ment. As a result, the internal forward model enables thebrain to execute the movement with increasing precision,absent the need to refer to feedback from the moving bodypart. The cerebellum learns through practice to performoperations faster and more accurately, which explains howa person is able to move more skillfully and automaticallyafter repeated practice [42].

In this way, a forward model can be considered apredictor of the consequences or outcomes of motorcommands, which allows the behavior to become increas-ingly precise and automatic and allow it increasingly to beexecuted by processes outside of conscious awareness. Thefeedback limb of cerebro-cerebellar circuitry projects themost efficient behavior to motor regions of the cortex,allowing the cortex to store the most efficient representationof the behavior [26, 78, 79]. In other words, the cortex“stores” the actual procedural memory, having been“taught” or instructed by the cerebellar model.

When the behavior is automatic, one way to release it isthrough a voluntary motor command. But an individual canbecome so skillful and can acquire such expertise that he orshe no longer needs to be conscious of the intention toinitiate a movement. In a professional basketball or soccergame, the participants engage in multiple rapid, precise,creative movement sequences, each aspect of which theynever could have practiced for each specific context inwhich they are generated. After the game, when the athletewatches a recording of the competition, he/she recalls little

Cerebellum

or nothing about how the artistic movements wereexecuted, and may not even remember the movementsthemselves. This implies that they were performed auto-matically, without conscious awareness. These movementsoccur creatively, “on the fly” and “on-line” as the situationdevelops and unfolds, absent the need for consciousdirection. While aspects of these behavioral sequences havebeen interpreted as a combination of forward models,cerebellum inverse control models also appear to benecessary in the development of higher levels of skillautomatization [80].

Inverse cerebellar models transform intended actions orgoals into motor commands unconsciously, in order toreach those goals [31]. In other words, inverse modelsgenerate motor commands, helping, or even by-passing theprimary motor cortex [49]. At the risk of oversimplification,a forward model can be understood as a predictor. Aninverse model can be understood as a controller. (While acomprehensive review of cerebellar forward and inversecontrol models is beyond the scope of this paper, the readeris referred to Ito 2005 and 2011; and Kawato et al. 1987 foradditional information). In following sections, we will addresshow cerebellar control models are acquired and localizedwithin the brain, how these models are represented withinspecific cerebellar modules, and how switching from onecontrol model to another without conscious awareness isnecessary to achieve ongoing sensorimotor adaptation. Thisprocess represents an anatomic underpinning for conceptssuch as EF and ‘embodied cognition.’ These concepts, whichdirectly link movement to thought, will provide thefoundation for understanding adult behavior as well aspediatric neurodevelopment.

Features and Characteristics of Control Models

As is the case for the cortex, the human cerebellummaintains a functional topography, so that different cere-bellar regions or zones contribute to different behaviors[81–84]. Functions are represented asymmetrically [85].The cerebellum is generally organized along anterior–posterior and medial–lateral gradients. Because the prefron-tal regions from which cerebellar circuits originate areinvolved in cognitive functioning and in movement, theoutputs of these circuits provide the cerebellum with theanatomical substrate to influence the control of both [70,86]. Motor information is primarily represented in theanterior lobes, while cognitive information is represented inthe posterior and inferior lobes. Limbic information isrepresented medially, and cognitive information processingis represented laterally. The deep cerebellar nuclei, andspecifically the dentate nucleus, also appear to be topo-graphically organized. Dum and Strick (2003) have

proposed a topographic organization within the dentatenucleus of the monkey, and fMRI data demonstrate that thehuman dentate nucleus is also subdivided into a rostral andmore dorsal motor domain and a ventrocaudal non-motordomain [87–89]. These data are particularly significant inrelationship to topographically represented and modularlyorganized cerebellar control models.

Control models have primarily been investigated throughstudies of tool use. Tool use has a prominent and lengthyhistory in the anthropologic and phylogenetic study ofprimates, including sapiens, primarily because levels ofcognitive development are reflected by the relative sophis-tication in development and application of tools [90–93].Tools can be simple or extremely complex. The skillfulmanipulation of tools can be considered as extensions ofthe limbs, therefore reflecting action control. The tools weuse might even be considered as newly added parts of thebody. The ability to make tools implies ‘EF’ in terms ofplanning, the ‘off-line’ simulation or imagining the materi-als necessary to construct tools, the sequential processingnecessary for tool assembly, and the motor dexterityrequired for tool construction and eventual use or applica-tion. Tool use requires goal directed action in the service ofrealizing intentions. We all use tools today, ranging from‘mechanical’ tools, to writing implements, silverware,bicycles, automobiles, wheelchairs and even crutches, tomodern technologies such as computers and “smartphones.” Our adaptation is dependent on our ability to usetools effectively.

Complex tool use requires semantic/declarativeknowledge about familiar tools and their uses, as wellas the procedural skills we acquire that are necessary toperform the proper actions with these tools. At thecortical level, these requirements depend upon function-ally specialized networks involving temporal, parietal,and frontal areas within the left cerebral hemisphere,(which we described as the dorsal information processingpathway) [94]. At the most basic level, according tofunctional imaging data, the temporal cortex supportstool identification, while the parietal–frontal mechanismssupport tool use skills. While Krienen and Bucknerrecently identified four separate frontal–cerebellarcircuits, of particular interest were the additional findingsthat two posterior regions of the lateral cerebellum(regions that are characterized by secondary sensorimotorfunctions of the cerebellum) also projected to lateralparietal and temporal lobe regions [95]. Schmahmannand Pandya have previously identified reciprocal superiorand inferior parietal and superior temporal cerebro-cerebellar connections as well [64]. This establishes afunctional ensemble or network that can be considerednecessary to support rapid and skillful tool use andmanipulation.

Cerebellum

Realizing intentions and reaching goals accurately,precisely, rapidly and flexibly requires the brain to learnhow to control its action on external objects (tools) and topredict or anticipate the consequences of this control.These processes require us to mimic the input–outputproperties of our own bodies as well as the properties ofother objects. In other words, they require us to anticipatesensorimotor feedback. These processes define the charac-teristics of cerebellar control models. The above describedcircuitry establishes a framework within which thesemodels operate for skilled tool use.

A variety of fMRI studies have focused upon cerebellaractivity when human subjects were required to learn to usea novel tool. Many of these studies have been performedand/or summarized by Imamizu and Kawato [31, 32, 96].These studies have consistently demonstrated significantactivity in the lateral cerebellum after learning occurs. It hasbeen concluded that this region of the cerebellum is at leastone of the areas where internal control models thatrepresent the sensorimotor properties of controlled objectsare acquired. We believe this finding may be of additionalsignificance because it implies that this region is involvedin the synthesis of procedural with declarative information.The synthesis of procedural with declarative information, infact, may be an additional property of cerebellar controlmodels that should be articulated, particularly in view ofthe manner in which control models are believed to develop(see above). This issue is clearly relevant to furtherconceptualizing the cerebellum's role in thinking (see Ito[42, 49]; Vandervert et al. [174]; Koziol et al. [3]). Whenwe think during problem solving, we are usually referringto declarative information.

While the modular architecture of the phylogeneticallyolder parts of the cerebellum is known, Imamizu andcolleagues investigated whether or not a comparablemodularity exists in the lateral cerebellum for cognitivefunctions [97]. Using fMRI after subjects intensivelylearned to manipulate two novel tools, they found thatlateral and posterior cerebellar activation for the twodifferent tools was spatially segregated, with less than10% overlap. They also found that subjects could easilyswitch between using these two novel tools. They inter-preted these findings as reflecting modularity within thelateral posterior cerebellum for two different cognitivefunctions. Higuchi, Imamizu, and Kawato assessed cere-bellar activation when subjects used sixteen very commontools (such as scissors, chopsticks, and a hammer) with theassumption that use of these instruments was automatic[98]. They compared this to cerebellar activation when thesubjects only mentally imagined using these same toolswithout actual hand movements. Activation during actualtool use was primarily located in the anterior lobe of thecerebellum. By contrast, activation during imaginary tool

use was localized laterally, in the posterior lobe of thecerebellum. For each tool in the imaginary condition, theyalso measured the distance of the area of activation from thefourth ventricle and found that imagined tool use wasrepresented in various different posterior lateral areas. Theyinterpreted anterior lobe activation as reflecting the senso-rimotor activity of the limb muscles, while they interpretedactivation within the various regions of the posterior lobesas reflecting internal models for tool use.

The results imply that internal models for the skillful useof common tools are organized in a modular fashion, withdifferent regions of the lateral cerebellum contributing tothe use of different tools, and they imply that purposive,“cognitive” information about objects is modularly repre-sented within the cerebellum. As noted by Ito, it isgenerally accepted that the operations and “events of thecerebellum are not reflected in consciousness, in which thecerebral cortex has been generally involved” (page 101)[77]. These findings thus also imply that this information isoutside of conscious awareness. In our view, this modularorganization has implications for understanding the neuro-anatomy of embodied cognition, which will be discussed ina following section.

Rapid and efficient sensorimotor interaction with theenvironment is an inherent requirement for successfuladaptation. This implies that humans can switch from oneinternal model to another based upon contextual informa-tion. The cerebellum links a behavioral context to a motorresponse [61]. This is one of the inherent properties of acerebellar control model. After practice, when the behavioris automated, the occurrence of the context can trigger theresponse. When context changes, behavior changes throughinteractions of forward and inverse models. However, notall behavioral switching is the same. Imamizu and Kawatohave identified neural correlates of predictive and post-dictive switching mechanisms for internal models [99].These identified switching mechanisms have somewhatdifferent neuroanatomies, which has implications forpractical behaviors.

Earlier, we described the functional neuroanatomy of thedorsal information processing stream and its significance insensorimotor interaction. Activation in the dorsolateralprefrontal cortex, the insula, the anterior aspects of theintra-parietal regions, and the lateral cerebellum are relatedto the switching of internal models [100]. Internal modelsplay an important role in switching mechanisms within theparietal and cerebellar regions. Two types of informationappear to be critical for switching internal models. Onetype, contextual information, can be perceived beforemovement execution. The other type, information aboutthe difference between actual and predicted sensorimotorfeedback, is computed either during or after movementexecution. Contextual data, which is essentially information

Cerebellum

known beforehand, allows for predictive switching; thisrecruits activation within the superior parietal lobule.However, when switching is dependent upon predictionerror for postdictive switching, activation is observedwithin the inferior parietal lobe and prefrontal cortex. Inthese ways, there are regional differences in the neuralsubstrates between predictive and postdictive internalmodel switching mechanisms at the level of frontal–parietalinteractions in combination with cortico-pontine cerebro-cerebellar circuitry [31].

Dias-Ferreria, Sousa, and Costa reviewed studies thatidentify ways in which the frontal cortices are connected tothe olivary system and serve as additional, alternativepathways to the well known cortico-pontine projectionsystem [64, 67, 101]. It has been proposed that these fronto-cerebellar (olivary system) connections are involved inswitching from voluntary to completely automatizedbehaviors during skill learning. These connections addi-tionally provide automatic sensorimotor surveillance duringthe performance of actions or behavioral sequences in anincreasingly precise and accurate manner. These regionshave all been implicated in higher-order functions requiringstimulus-outcome encoding and automation of recurrentbehaviors, while they also provide additional circuitriesthrough which control models can use stimulus-responsecontextual coupling to initiate automatized behavior.

These switching mechanisms have practical implicationsto the degree that they influence behavioral adjustmentspeed and the ability to make transitions in a constantlychanging environment. In this way, these mechanismswould appear to be related to the “executive control” ofbehavior. These switching mechanisms are also essential toour initial description of a behavioral model based uponongoing sensorimotor interaction with the environment.Similarly, these switching mechanisms have implicationsfor pathological behaviors. For example, parents of childrenwith neurodevelopmental disorders often describe theiroffspring (children and young adult) as having difficultiesin making transitions. These children rely on routine, oftento the extent of becoming “oppositional.” This presentingproblem also correlates with aspects of the recentlyidentified cerebro-cerebellar circuitry involved in differentpresentations of Attention Deficit Hyperactivity Disorder[102, 103]. This “switching” circuitry may be dysfunctionalin children with this described behavioral rigidity.

Embodied Cognition and the Cerebellum

A school of thought in cognitive psychology focuses uponthe concept of “embodied cognition,” which seeks to placecognition within a sensorimotor context. However, whilethis concept of embodiment has become increasingly

popular, there is limited agreement on what the term“embodied cognition” exactly means and to what extent“embodiment” includes sensorimotor versus higher levelcognitive function. Svensson and Ziemke [104] argue thatthe key to understanding the embodiment of cognition isthe sharing of neural mechanisms between sensorimotorprocesses and higher-level cognitive processes. However,just what exactly is a “higher-level” cognitive process? AsSaling and Phillips [30] have pointed out, and we agree,automatic behaviors are efficient and not mindless. Theyare essential to “executive functioning” in that they servethe best interest of the organism as a whole for the purposeof survival, and yet they are not typically considered“higher-level” processes.

Similarly, sensorimotor and procedural learning systemsand “higher level cognitive” learning and memory systemshave traditionally been understood as supported by distinct,separate neuroanatomies involving the neostriatal systemand the medial temporal lobe (hippocampal) system,respectively. While the brain's memory systems operate inparallel to support behavior, how these procedural anddeclarative memory systems interact is not fully known andcontradictory findings have been reported [105, 106].However, when considering the role of interactions oflearning and memory systems in embodied cognition, thecomputational model originally proposed by Doya is ofcritical significance [107]. With respect to a “globalarchitecture” of ongoing adaptive behavior, the “super-vised” learning modules (internal models) of the cerebellumare reciprocally connected with body and movementschemas stored in parietal and frontal cortices (cerebro-cortical models). The basal ganglia's instrumental learningsystem, essentially a reinforcement-based learning module,functions to evaluate the given state/situation while selectingan action, stored in the cortex, based upon the outcomevalence of that evaluation.

Doya's computational work originally conceptualized thedeclarative/episodic “unsupervised” learning modules ofthe cerebral cortex as the medium for representing theconditions of the external world as well as the internalcontext of the individual. The neocortex was characterizedas providing the common representational basis for thecerebellum and the basal ganglia “between which there areno direct anatomical connections” (page 966). Nearly theentire neocortex projects to the neostriatum (with theexception of the posterior regions of the occipital lobes),while the cortex similarly projects to the cerebellum (withthe exception of the inferior temporal cortex). While it iscertainly true that this organizational profile ensures that thecerebral cortex, basal ganglia, and cerebellum are allworking with the same data or information, it is nowknown that the basal ganglia and the cerebellum arereciprocally connected.

Cerebellum

The cerebellar dentate nucleus projects directly to thestriatum, which is the primary input region of the basalganglia [108]. The subthalamic nucleus also projects to thecerebellar cortex [109]. Therefore, the basal ganglia and thecerebellum communicate with each other. In addition, it hasrecently been demonstrated that the cerebellar dentatenucleus modulates dopamine release in the medial prefron-tal cortex, which plays a role in behavioral outcomeprediction and can influence the instrumental reinforcementlearning within the basal ganglia [110]. It remains to behypothesized and demonstrated how these pathways mightfunction in the brain's learning system interactions. Ourdiscussion is presented solely to better establish a theoret-ical groundwork or hardware for procedural and declarativememory interactions that are fundamental to definitions of“embodied cognition.”

Emulation/Simulation and Learning

It is well known that mental rehearsal or imagining theexecution of a motor activity (which is essentially the off-line simulation of action that can be described throughdeclarative recall) improves motor performance [111]. It isalso documented that similar brain regions are recruited andactivated during the performance of an activity and whenimagining doing it [112]. Wadsworth and Kana [113] foundthat viewing tools seems to generate prehensions andanticipations about using them, and that the imaginedaction of using a tool mirrored brain responses underlyingthe functional use of the tool. The authors found thatperception and contextual coupling along with imaginationof tool use appeared to generate precursors to overt actioncontrol over tools. Similarly, it has been found thatplanning the use of a tool activates a distributed networkin the left hemisphere. This network consists of theposterior superior temporal sulcus, and proximal regionsof the middle and superior temporal gyri; inferior frontaland ventral premotor cortices; parietal areas including theanterior and posterior supramarginal gyrus; and the dorso-lateral prefrontal cortex. It should be noted that theseregions include what has been described as the “mirrorneuron” system [49, 114, 115]. With the exception of theleft DLPFC, adjacent and partially overlapping regions ofleft parietal, frontal, and temporal cortex are engagedduring action execution. This entire lateralized networkcomprises a functional ensemble for the interaction ofsemantic and motoric representations upon which mean-ingful skills depend [116]. (This complex network waspreviously illustrated in Fig. 1.) Declarative knowledge canthus facilitate procedural learning.

So, imagine that you wanted to learn how to use a newtool, and that until now the only experience you had using

the tool was through observation and your own practice viamental imagery. You would already “know” how to use thetool. Your observations and imagery actually representsemantic, declarative knowledge; this is information aboutgripping and moving/turning the tool that you could“declare” through putting in words. This information isretained in posterior cortices in interaction with the medialtemporal lobe memory system. In order to improve yourperformance, you practice repetitively. You control “action”by incorporating “declarative knowledge” (which is whatyou know about the tool) to guide your behavior through“executive control.” With continued practice, your use ofthe tool improves. Your tool use eventually becomes rapid,flexible, and automatic. You have “learned” more about theinput–output properties of the tool (its sensorimotorproperties) and how it works (its dynamic properties) byactually using it. Your use of the tool has become automaticand you no longer think about the “executive control”guidance involved in first using the tool.

So what has occurred here? You have constructed andmodified cerebellar and cortical internal control models,and the behavioral product includes both procedural anddeclarative information. Your procedural and declarativeknowledge have been grounded in sensorimotor anticipa-tion [20]. In this way, you have used declarative knowledgeto develop and modify a “new” behavior. The improvementin your performance reveals that your unconscious“on-line” sensorimotor predictions were correct, while yoursemantic declarative knowledge about the tool and how touse it remain available for “off-line” simulation (or thought)that you could communicate or “declare” to another person.In this way, your overall knowledge of the tool has becomean “embodied cognition” through the operations of controlmodels. The cerebellar control model allows for theautomatic, nonconscious application of tool use knowledge,while declarative, semantic, cortically represented informa-tion about the tool's use remains available for “off-line”simulation and planning. This interaction between proce-dural and declarative/semantic memory systems is at theheart of “embodied cognition.” To the extent that cerebellarcontrol models assist in the development of procedurallearning, the cerebellum would be involved in “embodiedcognition,” as is suggested by the tool use studies byImamizu, et al. referenced above.

Our example illustrates how unconscious cognition canguide “automatic” EF behavior to the extent that use of thetool is adaptive to the situation and serves the individual'sbest interests at the time. We believe this viewpointchallenges traditional “perception-cognition-action” behav-ioral control paradigms while it allows for the concept thatunconscious cognition guides adaptive executive function.However, the theoretical underpinning of our example hasnot yet been demonstrated in an evidentiary way and

Cerebellum

warrants investigation at an experimental level. Further,while our example might “work” to explain adult-level“embodied cognition,” at first glance, the proposal may beproblematic for the pediatric population, which is the topicfor discussion below.

Pediatric Neurodevelopment

Humans are not born with well-developed sensory andmotor systems, nor are they born with well-developedmedial temporal lobe memory systems upon which declar-ative (semantic and episodic) memories depend. This isillustrated by the phenomenon of “infantile amnesia,”which refers to the fact that we are unable consciously torecollect events or experiences from the first 2 to 3 years ofour lives [117]. The phenomenon of “childhood amnesia”refers to the period from 3 to 5 years in which more event(episodic) memories are accessible to adults, but thesememories are lacking in specific detail and center aroundglobal feelings of familiarity and a few facts [118]. Infantileand childhood amnesia are not simply manifestations ofhippocampal and/or temporal lobe immaturity. Rather,functional immaturity of the association and prefrontalareas of the neocortex represent the main limiting factor[119]. Semantic memory (memory for facts) and episodicmemory (memory for events/experiences of one's life) aredevelopmentally dissociable. Raj and Bell [120] note thatchildren can acquire semantic information as early as thefirst year of life, demonstrated by increasing vocabularyand language skills, while their episodic memory system,which binds facts, events, and experiences and is moredependent upon prefrontal cortex, develops later. Hayneand Imuta conclude that by the age of 3 years, childrenexhibit rudimentary episodic memory skills. The semanticmemory system develops first, which appears to providestructure and meaning to the episodic memory system.Similarly, procedural memory systems are functionally“on-line” before declarative memory systems, while thetwo systems (declarative and procedural) are developmen-tally dissociable and follow different developmental trajec-tories [119]. We consider the how these features of childfunctioning relate to information acquisition below.

The Development of Motor Skills and EF

Over the past decade or so, considerable interest hasemerged concerning the relationship between motor skilldevelopment and the subsequent development of cognition,and particularly executive function. In infancy, suckingskills predict neurodevelopmental outcomes; babies whoexhibit poor suck coordination are likely to experience

complex motor and cognitive delays later [121]. Suck,feeding, and the production of speech and language areencoded and modified by overlapping neural networks ofcortical, subcortical, and brainstem regions that support“executive control” later. Piek and Dawson et al. [122] havedemonstrated a strong relationship between early grossmotor problems and the later development of particular EFdeficits in processing speed and working memory(as assessed by the WISC-IV) in school-aged children. Inanother study, Westendorp et al. [123] compared 7- to12-year-old children with learning disabilities to age-matched normal controls on measures of locomotor skillsand object-control skills as well as reading and mathematicstasks. Children with learning disabilities performed morepoorly on all motor tasks. Additionally, a relationship wasobserved between reading and locomotor skills andbetween mathematics and object control skills. A largerlag in specific learning skills was accompanied by poorerspecific motor skill scores. Many investigators have beenpuzzled by this apparent link between early motor and latercognitive, EF development. Why does motor developmentpredict the development of executive cognitive ability andacademic skills?

We are well aware that there can be considerablevariability in neuromotor development from kindergartenthrough adolescence [124]. Nevertheless, we believe thatunderstanding the relationship between motor developmentand EF is deceptively simple. The brain did not evolve forthe specific or explicit purpose of developing cognition.While the clinical scientific focus has been upon higher-order cognitive phenomena, the fact that perceptual andmotor capacities developed first, and dramatically, as wehave ascended the phylogenetic scale has been ignored.

We were not born to think. We were born to move.Human creative ideas are nothing in the absence of themanual dexterity that allows tools to be made, complexarchitecture to be constructed, art to be created, andinstruments to be played. Communication of complexinformation and ideas is accomplished through abilitiesthat require complex motor skills such as gesture, speech,writing, or typing [125]. Humans can make precise, highlydifferentiated movements, can combine individual move-ments into high-order sequences that can be simulated orthought about, and can execute movements and their uniquecombinations rapidly and automatically, even withoutconscious thought. It is no biologic accident that the bodyparts with the greatest sensory sensitivity, motor activity,and dexterity have the largest regions of representationwithin the brain [126].

The fundamental purpose of an organism and a person isto survive through environmental interaction. This requiresongoing sensorimotor learning within a changing environ-ment. Movements are not random. They are organized and

Cerebellum

purposeful, even during infancy and childhood [127, 128].Lockman [129] has characterized tool use as a continuousdevelopmental achievement that is rooted in the perception-action routines that infants employ to gain informationabout their environments. Infant pointing is meaningfulmovement. Tomasello, Carpenter, and Liszkowski havedemonstrated that when 12-month-old infants point for anadult, they are in some sense trying to influence the adult'smental state [130]. By 3 years of age, toddlers do notindiscriminately imitate actions with a tool. Rather, theyselectively reproduce actions that have a desired causaleffect [131]. Therefore, it is essential to consider why amovement is made, in addition to how movements areplanned, and how a movement anticipates what is going tohappen next. These features of movement require EF, ifonly on a rudimentary level.

The development of goal-directed action managementrequired development of anticipatory control mechanismsin order to predict the sensorimotor outcomes of action in“real time” as situations develop. This resulted in thedevelopment of mechanisms for “on-line” sensorimotoranticipation in order to adjust to the environment “on thefly” and “off-line” simulations of possible potential actionsin order to generate purposive, intentional, goal-directedbehaviors in the future [52]. These “off-line” simulationsinclude “executive function,” which is essentially anextension of the motor control system. Therefore, move-ments in the normally developing child should never beconsidered random; they are always purposive and accom-plish goals.

In this way, motor development and EF are inexorablyand inextricably linked because deviations/abnormalities inmotor development represent an early failure in “actioncontrol” which is essentially an early form of “EF.”Instability in the development of the motor system impactsupon the ability to develop “cognitive control” over motorsystems. The motor system allows cognition to evolve inorder to develop simulative, “off-line,” cognitive controlcapacities. We believe that deficits within the motor system,as might be manifest by abnormalities or inconsistencies inmotor development, generate (or at the very least, predict)developmental disorders. This is a viewpoint that is sharedby many other researchers [127, 128, 132–134].

Within human development, we believe these controlmechanisms proceed in a predictable way, and they allowfor overlap as the behavioral repertoire expands. First, thereis movement (even if only manifest by the seeminglysimple yet goal-directed suck reflex). Next, movement andcognition are coincident. As we move, we detect andobserve movements of objects, and in this way, we discoverobjects and learn about their properties. Subsequently,movement and cognition coexist. We might move to obtainsomething we perceive or move in response to something

we perceive, or we might move to acquire a new skill, suchas learning to dress by using buttons and zippers, learningto tie shoes, or learning to use silverware. Finally, cognitionbecomes subordinate to movement, as behavioral automa-ticity is achieved, and the “off-line” executive simulation ofactivity becomes necessary only for novel problem solving.In this way, we can directly link movement to thought, andcan understand “EF” as an extension of the motor controlsystem [3, 135, 136].

Movement Control in Childhood: NeuroanatomicCandidates

The brain regions and mechanisms controlling complexmovement and “EF” in adults are in process of becomingwell understood [26, 30–32, 39, 40, 137–141]. Develop-ment of movement programs is under the influence offrontal lobes (including the frontal eye fields) and posteriorparietal interactions. Program execution is presumably theresult of frontal–basal ganglia interactions. The superiortemporal sulcus (STS) is recruited in detection and incertain types of movement guidance. The “EF” of “workingmemory” that guides new behavior appears to operate onthe basis of a parallel cortico-basal ganglia interactionmodel, while the learning of procedures and their adapta-tion across contexts appears to be under the influence of thecerebellum.

Considerably less is known about the development ofneuroanatomic control mechanisms and processes in chil-dren. A child slowly develops within an environment thatdemands ongoing sensorimotor interaction. Before infantsmaster the ability to reach and grasp objects, theyrepeatedly try to move their hands to the objects, and theypersist despite numerous failures. Children often try to walkvery early, when they can locomote more efficiently bycrawling [133]. While we fully appreciate that motivation isimportant in generating behavior, in these examples, thereis no obvious “reward,” and in this regard, we currentlyhave limited knowledge about how the reward systemdevelops aside from the rewards that are linked toattachment mechanisms (the discussion of which is outsidethe scope of this paper) [142].

Despite the fact that activity is so central to childdevelopment, the cerebellum is rarely considered withinthe context of motor development. We know that thevestibulocerebellum is believed to be mature at the time offull-term birth, while studies of newborn babies and infantsreveal that reciprocal cerebellar–vestibular interconnectionsare already fully myelinated [143, 144]. Cerebro-cerebellarcircuitry very closely resembles adult circuitry during thecourse of development [145], while by the age of 4 years,ponto-cerebellar tracts attain a staining of adult type/

Cerebellum

resemblance [146]. In addition, at birth, the cerebellarcortex exhibits its well-established architecture. All thePurkinje cells are present, as are the climbing fibers thatsynapse with the cell bodies [147]. Cerebellar adult volumecan be achieved between the ages of 7 to 11 years, withpeak size typically reached at the age of 11.3 years infemales and 15.6 years in boys [145, 148]. Superior motordevelopment in typically developing right-handed childrenis dependent upon the establishment of neocortical left-hemisphere dominance in intrinsic motor network connec-tivity. This includes frontal motor cortices, basal ganglia,and thalamic connections. Supplementary motor–anteriorcerebellar connectivity is included in this circuitry profile[149].

The cerebellum is also important because cerebellardevelopment is affected by premature birth. Premature birthis a common enough occurrence, and even relatively mildprematurity, defined within the range of 37 to 38 weeks canbe associated with structural abnormalities within focalcerebellar brain regions [150]. There is an abundant clinicalliterature which demonstrates that preterm birth is associ-ated with a high incidence of cognitive disorders [151,152]. The cerebellum's growth and development are rapidduring late gestation, but this growth is impeded bypremature birth [153]. Very preterm birth, defined as lessthan 32 weeks gestation, is associated with three possiblepatterns of abnormality in cerebellar development. First,there can be volume reduction of the cerebellar hemispheresand a smaller vermis. Second, there can be volumetriccerebellar hemispheric reduction with an enlarged fourthventricle and a deformed vermis. Third, there can be normalcerebellar shape but with extensive reduction in itsdimensions [154]. These types of abnormalities have beenassociated with a variety of EF, cognitive, and motordifficulties [155–158]. These EF, action control deficits areoften chronic [159, 160]. Therefore, it is clear thatperturbations in cerebellum development can often resultin “cognitive” deficits. These deficits, and their persistence,may result from the fact that preterm delivery disrupts thedevelopmental program of the cerebellum [161].

Control Models in Children

Within this general context, it is possible to hypothesizehow the cerebellum might develop internal control models,particularly with respect to tool use. Tools—such assilverware, which children must learn to use and master—can be considered “extensions of the limbs” that “enhancethe efficiency with which skilled actions are performed”[162]. The use of silverware in children has obviousimplications for the development of anticipatory controlmodels, particularly since a child cannot “see” her mouth

while learning to use the tool (this process mimics thefinger-to-nose test that requires a cerebellar control model;Ito [49], pg. 169). Tool use in children is heavily dependentupon dorsal information processing pathway developmentand integrity. The “visual control area for actions” consistsof reciprocal premotor frontal-posterior parietal lobe con-nections, connections between the parietal lobe and thefrontal eye fields, connections between the frontal eyefields and the cerebellum, and the connectional profiles ofseparate frontal-cerebellar circuits, including the motorcircuit and the two circuits that associate segregated regionsof the lateral cerebellum with lateral parietal and STStemporal lobe regions [54, 95, 163–165]. In our discussion,we again focus on the possible role of the cerebellum.

By the time children are 3 years old, most use an adult,radial clenched grip for silverware, with the thumb pointingtowards the head of the tool, although before that time (andafterwards in some children), children experiment with avariety of grip patterns before they develop a grippreference [132]. Based upon imaging studies of tool usein adults, we would expect that the anterior region of thecerebellum would be one of the areas activated forsilverware grip. The movements used for bringing food tothe mouth using a spoon and using a fork are somewhatdifferent, while the grip and movement of the knife, whichis a tool use that is developed later, is different frommovements required for the spoon and fork. Yet withrepeated practice, as months and perhaps years go by, thenormally developing child develops mastery, automaticity,and flexibility with all of these grips and related egocentricmovements.

Based upon brain imaging investigations, we wouldpredict that the characteristic movements required for eachpiece of silverware would be represented in proximallyclose, yet distinct regions of the modularly organized lateralposterior cerebellum [98]. If we extrapolate from the adultevidence, we would have little reason to suspect that thatthe localization of silverware use would reorganize withinthe cerebellum once the use of these “tools” was acquiredthrough the development of models (body schema, motorschema, and cerebellar internal control models).

We have used this example to illustrate severalpoints. First, a complex and widespread network ofbrain regions would initially be necessary to acquire thebehaviors. Second, because the frontal cortices andcerebellar regions to which they project develop rela-tively slowly (though they just about keep pace witheach other) [166], we would expect these behaviors todevelop relatively slowly, even though the dorsal andventral streams are integrated by the age of 9 months [54].Lastly, and most importantly, the example illustrates thatprocedural learning can lead to and facilitate the devel-opment of declarative, semantic knowledge.

Cerebellum

Semantic memory, dependent upon the temporal andparietal lobes, frontal brain regions, and the basal ganglia,concerns knowledge that allows for the formation ofcategories, word meaning, and concepts [167, 168].Language can be considered a semantic system. It is the“great categorizer” that allows us to bring meaning to ourperceptions. Semantic categorization is an advantagebecause it allows information to be grouped conceptually,while similarities and differences can simultaneously bespecified to allow these concepts to be manipulated andapplied. For example, while we know that apples, bananas,strawberries, and oranges are conceptually similar becausethey are all fruits and can be substituted as nutrients, wealso know the properties of these fruits are different.Generalization and specificity both guide eating behaviorin relation to fruit, but it is also influenced by fruitpreference (reward valence) and availability. This processis subserved by sensory neocortical regions, the prefrontalcortex, the medial temporal lobe memory system, and thebasal ganglia and its dopaminergic reward system [167,169, 170].

In our procedural learning example, the child learns howto use tools efficiently within the semantic category ofsilverware, while also acquiring the ideas that a spoon doesnot serve the functions of a fork; that a fork, because of itsprongs, can never serve the function of a spoon; and thatwhile a knife is used for spreading and/or cutting food, itcan also be implemented for other functions outside of thecategory of silverware use. Silverware and dishware use aredirectly tied to the semantic, declarative category of foodtexture; a spoon is used for liquid, such as cereal or soup,which is served from a bowl; a fork is used for food withsolid texture, which can be placed on a plate or in a bowl,while the size and texture of meat or fish often necessitatesit be cut with a knife. A cup requires a different grip thandoes silverware, and even though the cup holds liquid, itscontent satisfies thirst, and not hunger; the category ofliquid, which encompasses both soup and juice, belongs tothe general semantic category of food, although thesenutrients serve different internal needs. By the time a childis 2 to 3 years old, when association cortices start to maturefor the purpose of storing semantic associations and whenlanguage begins to emerge, interaction with the environ-ment leads to exposure to different objects and “tools” thatare directly linked to separate and overlapping categories ofsemantic, declarative knowledge.

This grasp of categorical similarities and functionaldifferences, which has been acquired through sensorimotoranticipation inherent in the processes of eating anddrinking, is at the heart of declarative semantic knowledge.In other words, declarative, semantic knowledge, which isby definition available for “off-line” use (thinking andplanning), has actually been obtained through, and is

grounded in, direct sensorimotor interaction with theenvironment. In this way, the “EF” of eating and drinkingand the semantic conceptualizations acquired from partic-ipating in the activity all become “embodied cognitions.”They require the interaction of multiple brain networkswithin which the knowledge and behaviors are stored andlater manipulated. These processes begin with the firstmovements of childhood and are continuous throughoutdevelopment and adulthood.

While traditional theories of learning and memorytypically recognize procedural and declarative memorysystems as distinct, independent functions and processes,this sensorimotor learning model demonstrates that there isconsiderable overlap between them. Procedural and seman-tic knowledge are blended and become “embodied” asinformation is acquired through the brain's ability to gatherknowledge on the basis of its anticipatory and predictivecapacities. It is this procedural-declarative developmentallearning interaction that makes semantic informationavailable for “off-line” use in other cognitive processes,such as planning. The overlapping, widespread corticalnetwork, which includes temporal and parietal and parietal–frontal regions, provides the underpinning to store thissemantic, declarative information so it can be used for “off-line” conscious cognitive purposes. At the same time, thecerebellar control model allows for the application of theprocedures for silverware used in our example, with“cognition” falling outside of conscious cognitive aware-ness. While traditional theories of learning and memoryemphasize differences between procedural learning andsemantic memory systems, a sensorimotor learning modelnot only respects these differences, but it also revealsinteractions between these systems [52, 104]. The fact thatrecruitment in essentially the same networks is demonstratedin neuroimaging studies that assess perceptual/semantic tasksinvolving planning and executing activities related to tooluse helps to illustrate the neuroanatomic underpinnings of“embodied cognition” [116].

Discussion

We have presented a model of behavior that supports adual-tiered model of adaptation. This model is characterizedby automatic behaviors that alternate with episodes ofhigher-order decision making and control. While most dailybehaviors fall into the realm of automatic functioning withrapid “on-line” adjustments made through a process ofsensorimotor anticipation, information remains available for“off-line” use and manipulation for the purpose ofdeveloping intentions, plans, and goals. This paradigm isderived from a behavioral model that requires ongoinginteraction with the environment. From a neuroanatomic

Cerebellum

perspective, this model is based upon interactive brainnetworks that involve differentiated regions of cortex alongwith subcortical structures.

We have limited our discussion by emphasizing the roleof the cerebellum in only certain motor and non-motorprocesses. While cerebellar function can be interpretedwithin different schools of thought that emphasize itsvarying roles [171], we have focused upon a rapidlyemerging literature that stresses the cerebellum's ability toconstruct internal models for the unconscious control ofadaptive behaviors occurring in “real time.” These forwardand inverse models control behavior and cognition on thebasis of anticipation or prediction, as has been eloquentlydescribed by Ito [42]. We do not believe that the brainevolved for cognition per se; rather, the brain evolved forthe control of action or movement. This phylogeny requiredmechanisms that could automate and adjust behavior on thebasis of anticipated sensory feedback. At the same time,they could allow for the “off-line” manipulation of actioncontrol for the purpose of intentional, goal-directed behav-ior and novel problem solving for an individual interactingwith an ever-changing environment that called for flexibil-ity and innovative behaviors.

In describing the control of action, we have describedthe concept of “embodied cognition” which directly linksmovement to thought, and we have specified that “thought”may not necessarily be conscious in nature. For example,just as we typically “move” to accomplish goals withoutbeing consciously aware of the sensation of movement, wealso “think” without being consciously aware of “thinking.”This is made possible through cerebellar internal controlmodels. We have observed that this can be demonstrated inother species besides humans. Whether we see humanthought as “abstract” in any particular circumstancedepends upon our definition of abstract. From the brain'sperspective, thinking about doing something, such asmanipulating the necessary nuts, bolts, and other pieces inproper order to assemble a cabinet or bookcase, is not muchdifferent than performing the actual processes. This mightbe inferred from the imaging data we reported. As we haveindicated, both the processes of mental manipulation andactual object assembly recruit many of the same brainregions.

As Ito has indicated, once a to-be-controlled object iscoded within the cerebellum, movement, and thoughtbecome equivalent; the cerebellum will manipulate theinput in an identical way [4]. This type of manipulation, inthe absence of movement, is a form of abstract “thought.”We also recognize that prefrontal projecting areas of thecerebellar cortex process information that is of a purely“abstract” nature [73]. In our view, this reflects the modularorganization of the cerebellum and the dentate nucleus towhich its Purkinje cells project [89]. Throughout the course

of phylogeny, a critical role for the cerebellum has been toallow the brain to anticipate the outcome of sensorimotorbehavior. As the neocortex, including the frontal andprefrontal lobes expanded in parallel with the cerebellum,the basic functions, processes, and interactions remainedthe same. The abstract thought of prefrontal cortices servean extension of this system. The “data” stored andmanipulated within the motor regions of the frontal lobesprimarily consist of sensorimotor knowledge; the knowl-edge of the prefrontal cortex is based upon the manipu-lation of ideas. Just as the cerebellum informs motorregions about the most efficient way of executing behav-iors, it instructs the prefrontal cortex how to manipulateideas for problem solving. These brain regions do notfunction in isolation; they operate as a functional networkor ensemble.

Just as certain cerebellar modules are organized to thinkabout the sensory and movement properties of certainobjects and to anticipate how movement will be associatedwith a predicted outcome, Crus I and Crus II (HVIIA)appear to be modularly specialized for abstract thinking. Allthinking, or thought manipulation, includes the eventualparticipation in action control. This includes the abstractthinking required to compose this paper, which accom-plished the action-based potential for communication. Akey proposition is that the prefrontal cortex has no inherentability to “think ahead” but instead, learns to do so throughexperience (personal communication from E. Goldberg,April, 2011). What the cerebellum does for movement, itdoes for thought—including abstract thought. During thecourse of development, the inherent properties, character-istics, and functions of the cerebellum instruct or teach theprefrontal cortex how to “think ahead,” which is at the heartof abstract thought. This process mimics how cerebellarcontrol models instruct motor cortices to anticipate senso-rimotor, behavioral outcomes.

While ontogeny recapitulates phylogeny, this entiredevelopmental process occurs through a process ofsensorimotor learning. Thought is characterized byanticipation and prediction. The cerebellum plays a rolein driving the modulation of thought manipulation andspeed and accuracy outcomes, and it plays a moregeneral role in instructing the prefrontal cortex toexpect, predict, or anticipate outcomes throughout thecourse of neurodevelopment. While the motor cortices“think” about movements, the prefrontal cortex “thinks”about “thoughts.” The forward-focused prefrontal corti-ces of this paper's authors, for example, worked hard asthe authors manipulated the ideas necessary to developthese themes. The development and manipulation of thesethemes was grounded in the authors' anticipation orprediction of how the thought content would be inter-preted by the reader.

Cerebellum

The cerebellum is also a critical player in “EF,” whetherthis EF is an appropriate movement in “real time” or anappropriate, higher-order plan of action to be executed later.Moreover, understanding cerebellar contributions toabstract thought also help us understand how prefrontal–cerebellar interactions assist in solving novel problems,which forms the base of creativity and innovation. As Itorecently stated,

“It has been proposed that when a stimulus ispresented in the realm of thoughts and ideas, it ismaintained in the working memory, where it iscompared with mental models stored in the tempor-oparietal cortex. Comparison is also made to cerebel-lar internal models, which have been copied frommental models and stored in the cerebellum. If areceived stimulus matches a mental model or aninternal model, the stimulus is considered at the levelof conscious awareness to be familiar and the thoughtprocess is completed. If the stimulus is truly novel,however, existing models will not recognize thestimulus. Error signals are then generated that (1)suppress at the level of conscious awareness accep-tance of the idea that the mental model is familiar, (2)activate the attentional system, which in turn activatesthe working memory system, which will thenstrengthen the search for stored mental and internalmodels, and (3) modify internal models in thecerebellum if errors continue to occur. Eventually,the stimulus will find a matching counterpart in thepools of mental models and internal models. …Thefinal process is an “aha” at the level of consciousawareness; that is, expression of the fact that asuccessful conversion has taken place. In other words,the original stimulus is no longer novel. Thesepreceeding processes and mechanisms may well bethe basis for creativity and innovation (Ito 2011,pp197–198).” (This description also directly linksmovement to thought, emphasizing that what cerebellarmodels do for movement, they also do for thought).

We realize this paper has implications that go wellbeyond the scope of this manuscript. For example, althoughwe have mentioned how observation might enhancelearning, we have not examined how the “mirror neuron”system (which plays an important role in observationallearning) might contribute to the acquisition of proceduresand declarative, semantic concepts in other areas offunctioning. We have not fully explored implications forlanguage in particular, which would be a useful next step,particularly since Broca's area 44 presumably evolvedfrom—or is at least proximal to—what has been identifiedas a component of the frontal mirror neuron system. Theevolution of tool use and language have much in common

from an anatomic point of view [90, 91, 172–174]. Ullmanhas presented a model of language functioning that includesdissociable procedural and declarative memory systemsthat are dependent upon the same or at least similarbrain networks that subserve other functions, includingthose functions described in this paper [175, 176]. Whilethe cerebellum's role, with its forward and inverse models,has not been specifically investigated within Ullman's“declarative/procedural” model of language, we suspect itis involved in anticipatory processes associated withautomaticity.

Similarly, we have not specifically explored the clearimplications for social cognition and “theory of mind.”Glenberg and Gallese [177] have hypothesized that thesame premotor circuitry that controls action (adjacent to themirror motor neuron system) instantiates the “off-line”simulation of the observed actions of other people. Galleseand colleagues have hypothesized that the ability tounderstand other people's intentions relies primarily on themotor cognition that underlies one's own personal capacityto act [178]. Motor acts form the essential framework orbuilding blocks around which all activity is produced,perceived, and understood. Investigations of Autisticchildren reveal abnormal patterns of sensorimotor learningas well as impairment in the execution of “skilled” motorgestures, which very strongly correlates with impairment insocial and communicative functions. According to Mostof-sky and Ewen [179], this suggests that anomalous actionmodel formation, or abnormal cerebellar control models,may contribute to impaired development of social, com-municative, and motor capacities in autism. This clearlyimplicates the cerebellum in the formulation of a “theory ofmind.”

While the cerebellum can easily be envisioned inparticipating in so many functions, the question emergesas to whether or not it is capable of “running the wholeshow.” We believe the cerebellum plays a critical role inpediatric neurodevelopment because it participates in anetwork or ensemble of connected brain regions that drivevery broadly defined sensorimotor procedural learning and“action control.” This, in turn, indirectly contributes todevelopment of semantic/declarative systems that allow forthe “off-line” manipulation of information and simulativeprocesses such as planning. Over the course of develop-ment, this allows the prefrontal cortex to learn to “thinkahead” in higher-order control processes. As such, theability to anticipate is an inherent design characteristic of abrain that develops “bottom up.” However, sensoryperceptions are retained in the regions in which they areprocessed, while innervatory action/motor programs arestored within motor cortices. Cerebellar control models,which refine and adjust behaviors outside of consciousawareness, are stored within the cerebellum. As we have

Cerebellum

referenced above, the cerebellum, through its anticipatoryand correction functions, allows the cortex to retain themost efficient representation of a behavior. Without acortex, the cerebellum might be “useless” because it wouldhave no information upon which to base anticipatorycontrol. By contrast, without a cerebellum, the cerebrumremains capable of storing experience and motor programs.

Therefore, an adult with a history of normal adaptivefunctioning who experiences an acute onset condition, suchas stroke or tumor, is likely to demonstrate at least somerecovery (dependent upon extent and location of lesion).Developmental history establishes an experiential back-ground. While these patients generally do not achieveoptimal pre-morbid levels of functioning, they do attainsome improvement [180]. Functioning is less efficientwithout cerebellar input, but past experience retained incerebral cortex assists in adaptation. On the other hand, achild who never had much of a functional cerebellum tobegin with (cerebellar agenesis and posterior fossa tumorare just two beginning examples) experiences chronicadjustment deficits, depending upon the type of pathologyand the age of onset of the condition, which determines thelevel of cortical experience. Finally, certain psychiatricdisorders are characterized by cerebellar pathology. Affectedindividuals (as in autism and schizophrenia, to name just twoexamples) are chronically fed “bad data” from an abnormallyfunctioning cerebellum, which often results in chronicimpairment in EF. Therefore, within this interactive model inwhich brain networks operate as a functional ensemble, thereis no single brain region that can function independently in“running the whole show.” This makes both intuitive andphylogenetic sense.

While much of our argument is based upon a well-established neuroscientific knowledge base, we realize thataspects of this model, such as the developmental character-istics of memory systems and how different memorysystems might interact, is based upon minimal experimentalevidence. However, we have attempted to provide an“interim solution” to these issues while we have generateda framework for future consideration. We hope that thehypotheses we presented will contribute to future studies inthis field.

Summary

In this paper, we examined the close relationship betweenmovement and thought. In doing so, we rejected atraditional serial-order processing model of adaptation aswell as the phylogenetic notion that the brain evolved forthe explicit and specific purpose of “thinking.” Instead, wepresented a model of behavior based upon continuousinteraction with the environment in which behaviors occur

in “real time.”Within this model and the dual-tiered paradigmof adaptation it supports [135], we proposed that the brainevolved for the purpose of controlling action. This develop-ment required the evolution of both on-line (motor skillacquisition) and off-line (simulative thought rehearsal)sensorimotor prediction mechanisms. We focused on afundamental and comprehensive interactive role for cerebel-lar control models within these mechanisms. Why move-ments are made, how movements are planned and executed,and how movements might anticipate outcomes are ques-tions at the heart of understanding executive function andaction control, while the acquisition of movements play animportant role in defining aspects of embodied cognition. Wediscussed how this might be the case in both adults andchildren. By directly connecting movement to thought, wehave also examined important ways in what this modelmight apply to language processes, communication processesand social cognition, and we have suggested how the modelmight apply to concepts such as theory of mind.

Conflict of Interest Statement The authors have no conflicts ofinterest associated with this manuscript.

References

1. Cromwell HC, Panksepp J. Rethinking the cognitive revolutionfrom a neural perspective: how overuse/misuse of the term‘cognition’ and the neglect of affective controls in behavioralneuroscience could be delaying progress in understanding theBrainMind. Neurosci Biobehav Rev. 2011;35(9):2026–35.

2. van Schouwenburg MR, den Ouden HE, Cools R. The humanbasal ganglia modulate frontal-posterior connectivity duringattention shifting. J Neurosci. 2010;30(29):9910–8.

3. Koziol LF, Budding DE, Chidekel D. Adaptation, expertise, andgiftedness: towards an understanding of cortical, subcortical,and cerebellar network contributions. Cerebellum. 2010;9:499–529.

4. Ito M. Movement and thought: identical control mechanisms bythe cerebellum. Trends Neurosci. 1993;16(11):448–50.

5. Loring DW, Meador KJ. INS dictionary of neuropsychology.USA: Oxford University Press; 1999.

6. Lezak MD, Loring DW. Neuropsychological assessment. USA:Oxford University Press; 2004.

7. Gualtieri CT. The contribution of the frontal lobes to a theory ofpsychopathology. In: Ratey JJ, editor. Neuropsychiatry ofpersonality disorders. Cambridge: Blackwell Science; 1995.p. 149–171.

8. Dubois B, Pillon B, McKeith IG. Parkinson's disease with andwithout dementia and lewy body dementia. In: Miller EK,Cummings JL, editors. The human frontal lobes: functions anddisorders. New York: The Guilford Press; 2007. p. 472–504.

9. Miller R. A theory of the basal ganglia and their disorders. BocaRaton: CRC; 2008.

10. Haber SN, Rauch SL. Neurocircuitry: a window into thenetworks underlying neuropsychiatric disease. Neuropsycho-pharmacology. 2010;35(1):1–3.

11. Ardila A. On the evolutionary origins of executive functions.Brain Cognit. 2008;68(1):92–9.

Cerebellum

12. Zelazo PD, Qu L, Muller U. Hot and cool aspects of executivefunction: relations in early development. In: Schneider W,Schumann-Hengsteler R, Sodian B, editors. Young children'scognitive development: interrelationships among executive func-tioning, working memory, verbal ability, and theory of mind.Mahwah: Lawrence Erlbaum; 2005. p. 71–93.

13. Salloway S, Malloy P, Cummings JL. The neuropsychiatry oflimbic and subcortical disorders.Washington: American PsychiatricPress; 1997.

14. Frank MJ, O'Reilly RC, Curran T. When memory fails, intuitionreigns: midazolam enhances implicit inference in humans.Psychol Sci. 2006;17(8):700–7.

15. Wan X, Nakatani H, Ueno K, Asamizuya T, Cheng K, Tanaka K.The neural basis of intuitive best next-move generation in boardgame experts. Science. 2011;331(6015):341.

16. Manrique HM, Call J. Spontaneous use of tools as straws ingreat apes. Anim Cognit 2011;14(2):213–26.

17. Cantalupo C, Hopkins W. The cerebellum and its contribution tocomplex tasks in higher primates: a comparative perspective.Cortex. 2010;46(7):821–30.

18. Ramnani N. Frontal lobe and posterior parietal contributions tothe cortico-cerebellar system. Cerebellum. 2011. doi:10.1007/s12311-011-0272-3.

19. Pezzulo G. Coordinating with the future: the anticipatory natureof representation. Minds and Machines. 2008;18(2):179–225.

20. Pezzulo G. Grounding procedural and declarative knowledge insensorimotor anticipation. Mind Lang. 2011;26(1):78–114.

21. Piaget J, Inhelder B. The psychology of the child. 1972.22. Barsalou LW. Grounded cognition: past, present, and future. Top

Cogn Sci. 2010;2(3):322–7.23. Frank MJ, Doll BB, Oas-Terpstra J, Moreno F. Prefrontal and

striatal dopaminergic genes predict individual differences inexploration and exploitation. Nat Neurosci. 2009;12(8):1062–8.

24. Baddeley A. The central executive: a concept and somemisconceptions. J Int Neuropsychol Soc. 1998;4(5):523–6.

25. Isoda M, Hikosaka O. Cortico basal ganglia mechanisms forovercoming innate, habitual and motivational behaviors. Eur JNeurosci. 2011;33(11):2058–69.

26. Houk JC, Bastianen C, Fansler D, Fishbach A, Fraser D, ReberPJ, et al. Action selection and refinement in subcortical loopsthrough basal ganglia and cerebellum. Philos Trans R Soc LondB Biol Sci. 2007;362(1485):1573–83.

27. Bargh JA, Chartrand TL. The unbearable automaticity of being.Amer Psychol. 1999;54:462–79

28. Bargh JA. The automaticity of everyday life. Wyer, Robert SJr (Ed), et al. (1997). The automaticity of everyday life:Advances in social cognition, Vol. 10. (pp. 1-61). Mahwah,NJ, USA: Lawrence Erlbaum Associates, Inc. , Publishers.viii; 1997. 258 pp

29. Lakoff G, Johnson M. Philosophy in the flesh. 77th ed. NewYork: Basic Books; 1999.

30. Saling LL, Phillips JG. Automatic behaviour: efficient notmindless. Brain Res Bull. 2007;73(1–3):1–20.

31. Imamizu H, Kawato M. Brain mechanisms for predictive controlby switching internal models: implications for higher-ordercognitive functions. Psychol Res. 2009;73(4):527–44.

32. Imamizu H. Prediction of sensorimotor feedback from theefference copy of motor commands: a review of behavioral andfunctional neuroimaging studies. Jpn Psychol Res. 2010;52(2):107–20.

33. Shadmehr R, Smith MA, Krakauer JW. Error correction, sensoryprediction, and adaptation in motor control. Annu Rev Neurosci.2010;33:89–108.

34. Cisek P, Kalaska JF. Neural mechanisms for interacting with aworld full of action choices. Annu Rev Neurosci. 2010;33:269–98.

35. Shadlen MN, Movshon JA. Synchrony unbound: review acritical evaluation of the temporal binding hypothesis. Neuron.1999;24:67–77.

36. Singer W. Consciousness and the binding problem. Ann N YAcad Sci. 2001;929(1):123–46.

37. Stout D. The evolution of cognitive control. Top Cogn Sci.2010;2(4):614–30.

38. Bloedel JR, Bracha V. Duality of cerebellar motor and cognitivefunctions. Int Rev Neurobiol. 1997;41:613.

39. Hazy TE, Frank MJ, O'Reilly RC. Towards an executive withouta homunculus: computational models of the prefrontal cortex/basal ganglia system. Philos Trans R Soc Lond B Biol Sci.2007;362(1485):1601–13.

40. Frank MJ, Loughry B, O'Reilly RC. Interactions between frontalcortex and basal ganglia in working memory: a computationalmodel. Cognit Affect Behav Neurosci. 2001;1(2):137–60.

41. Stocco A, Lebiere C, Anderson JR. Conditional routing ofinformation to the cortex: a model of the basal ganglia's role incognitive coordination. Psychol Rev. 2010;117(2):541.

42. Ito M. Control of mental activities by internal models in thecerebellum. Nat Rev Neurosci. 2008;9(4):304–13.

43. Jacob F. Evolution and tinkering. Science. 1977;196(4295):1161–6.

44. Martin A. The representation of object concepts in the brain.Annu Rev Psychol. 2007;58:25–45.

45. Freeman JB, Dale R, Farmer TA. Hand in motion reveals mindin motion. Front Psychol. 2011;2:59.

46. Valyear KF, Chapman CS, Gallivan JP, Mark RS, Culham JC. Touse or to move: goal-set modulates priming when grasping realtools. Exp Brain Res 2011;212(1):125–42

47. Hendriks-Jansen H. Catching ourselves in the act: situatedactivity, interactive emergence, evolution, and human thought.Cambridge: MIT; 1996.

48. Freeman JB, Ambady N. A dynamic interactive theory of personconstrual. Psychol Rev. 2011;118(2):247.

49. Ito M. The cerebellum: brain for an implicit self. Upper SaddleRiver: FT Press; 2011.

50. Heilman KM, Rothi LJG. Apraxia. In: Heilman KM, ValensteinE, editors. Clinical neuropsychology. 4th ed. New York: OxfordUniversity Press; 2003. p. 215–35.

51. Barsalou LW. Grounded cognition. Annu Rev Psychol.2008;59:617–45.

52. Pezzulo G, Barsalou LW, Cangelosi A, Fischer MH, McRae K,Spivey MJ. The mechanics of embodiment: a dialog onembodiment and computational modeling. Front Psychol.2011;2:5.

53. Milner AD, Goodale MA. Two visual systems re-viewed.Neuropsychologia. 2008;46(3):774–85.

54. Njiokiktjien C. Developmental dyspraxias: assessment anddifferential diagnosis. In: Riva D, Njiokiktjien C, editors. Brainlesion localization and developmental functions. Montrouge:John Libbey Eurotext; 2010. p. 157–86.

55. Kaldy, Leslie AM. Identification of objects in 9 month oldinfants: integrating “what” and “where” information. Dev Sci.2003;6(3):360–73.

56. Doll BB, Frank MJ. The basal ganglia in reward and decisionmaking: computational models and empirical studies. Handbookof Reward and Decision Making. 2009;399:399–425.

57. Sheth SA, Abuelem T, Gale JT, Eskandar EN. Basal ganglianeurons dynamically facilitate exploration during associativelearning. J Neurosci. 2011;31(13):4878.

58. Heekeren HR, Wartenburger I, Marschner A, Mell T, VillringerA, Reischies FM. Role of ventral striatum in reward-baseddecision making. NeuroReport. 2007;18(10):951.

59. Haber SN, Kim KS, Mailly P, Calzavara R. Reward-relatedcortical inputs define a large striatal region in primates that

Cerebellum

interface with associative cortical connections, providing asubstrate for incentive-based learning. J Neurosci. 2006;26(32):8368–76.

60. Cisek P, Kalaska JF. Neural correlates of reaching decisions indorsal premotor cortex: specification of multiple directionchoices and final selection of action. Neuron. 2005;45(5):801–14.

61. Thach WT. Context-response linkage. In: Schmahmann J,editor. The cerebellum and cognition. San Diego: Academic;1997. p. 600–12.

62. Kinsbourne M, Jordan JS. Embodied anticipation: a neuro-developmental interpretation. Discourse Process. 2009;46(2):103–26.

63. Smaers JB, Steele J, Zilles K. Modeling the evolution of corticocerebellar systems in primates. Ann NY Acad Sci. 2011;1225(1):176–90.

64. Schmahmann JD, Pandya DN. The cerebrocerebellar system. IntRev Neurobiol. 1997;41:31–60.

65. Leiner HC, Leiner AL, Dow RS. Does the cerebellum contributeto mental skills? Behav Neurosci. 1986;100(4):443.

66. Balsters JH, Cussans E, Diedrichsen J, Phillips KA, Preuss TM,Rilling JK, et al. Evolution of the cerebellar cortex: the selectiveexpansion of prefrontal-projecting cerebellar lobules. Neuro-Image. 2010;49(3):2045–52.

67. Ramnani N, Behrens TE, Johansen-Berg H, Richter MC, PinskMA, Andersson JL, et al. The evolution of prefrontal inputs tothe cortico-pontine system: diffusion imaging evidence fromMacaque monkeys and humans. Cereb Cortex. 2006;16(6):811–8.

68. Middleton FA, Strick PL. Basal ganglia and cerebellar loops:motor and cognitive circuits. Brain Res Brain Res Rev. 2000;31(2–3):236–50.

69. Middleton FA, Strick PL. Cerebellar projections to the prefrontalcortex of the primate. J Neurosci. 2001;21(2):700–12.

70. Strick PL, Dum RP, Fiez JA. Cerebellum and nonmotor function.Annu Rev Neurosci. 2009;32:413–34.

71. Stoodley CJ. The cerebellum and cognition: evidence fromfunctional imaging studies. Cerebellum 2011 (in press)

72. Baillieux H, De Smet HJ, Paquier PF, De Deyn PP, Marien P.Cerebellar neurocognition: insights into the bottom of the brain.Clin Neurol Neurosurg. 2008;110(8):763–73.

73. Balsters JH, Ramnani N. Symbolic representations of action inthe human cerebellum. NeuroImage. 2008;43(2):388–98.

74. Ramnani N. The primate cortico-cerebellar system: anatomy andfunction. Nat Rev Neurosci. 2006;7(7):511–22.

75. Herculano-Houzel S. Coordinated scaling of cortical and cerebellarnumbers of neurons. Front Neuroanat. 2010;4:12.

76. Shadmehr R, de Xivry JJ Orban, Xu-Wilson M, Shih TY.Temporal discounting of reward and the cost of time in motorcontrol. J Neurosci. 2010;30(31):10507–16.

77. Ito M. Bases and implications of learning in the cerebellum—adaptive control and internal model mechanism. Prog Brain Res.2005;148:95–109.

78. Doyon J, Song AW, Karni A, Lalonde F, Adams MM, UngerleiderLG. Experience-dependent changes in cerebellar contributions tomotor sequence learning. Proc Natl Acad Sci USA. 2002;99(2):1017.

79. Galea JM, Vazquez A, Pasricha N, Orban de Xivry JJ, Celnik P.Dissociating the roles of the cerebellum and motor cortex duringadaptive learning: the motor cortex retains what the cerebellumlearns. Cerebr Cortex 2010;21(8):1761–70

80. Kawato M, Furukawa K, Suzuki R. A hierarchical neural-network model for control and learning of voluntary movement.Biol Cybern. 1987;57(3):169–85.

81. Stoodley CJ, Schmahmann JD. Functional topography in thehuman cerebellum: a meta-analysis of neuroimaging studies.NeuroImage. 2009;44(2):489–501.

82. Granziera C, Schmahmann JD, Hadjikhani N, Meyer H, MeuliR, Wedeen V, et al. Diffusion spectrum imaging shows thestructural basis of functional cerebellar circuits in the humancerebellum in vivo. PLoS One. 2009;4(4):e5101.

83. Glickstein M, Sultan F, Voogd J. Functional localization in thecerebellum. Cortex. 2011;47(1):59–80.

84. Stoodley CJ, Schmahmann JD. Evidence for topographicorganization in the cerebellum of motor control versus cognitiveand affective processing. Cortex. 2010;46(7):831–44.

85. Hu D, Shen H, Zhou Z. Functional asymmetry in the cerebellum:a brief review. Cerebellum. 2008;7(3):304–13.

86. Habas C. Functional imaging of the deep cerebellar nuclei: areview. Cerebellum. 2010;9(1):22–8.

87. Dum RP, Strick PL. An unfolded map of the cerebellar dentatenucleus and its projections to the cerebral cortex. J Neurophysiol.2003;89(1):634–9.

88. Kuper M, Dimitrova A, Thurling M, Maderwald S, Roths J,Elles HG, et al. Evidence for a motor and a non-motor domain inthe human dentate nucleus—an fMRI study. NeuroImage.2011;54(4):2612–22.

89. Thurling M, Kuper M, Stefanescu R, Maderwald S, GizewskiER, Ladd ME, et al. Activation of the dentate nucleus in a verbgeneration task: a 7T MRI study. Neuroimage 2011 (in press)uncorrected proof

90. Ambrose SH. Paleolithic technology and human evolution.Science. 2001;291(5509):1748.

91. Ambrose SH. Coevolution of composite-tool technology, con-structive memory, and language: implications for the evolutionof modern human behavior. Curr Anthropol. 2010;51(S1):S135–47.

92. Stout D, Chaminade T. Making tools and making sense:complex, intentional behaviour in human evolution. CambArchaeol J. 2009;19(01):85–96.

93. Langbroek M. Trees and ladders: a critique of the theory ofhuman cognitive and behavioural evolution in Palaeolithicarchaeology. Quaternary International, (in press) corrected proof

94. Johnson-Frey SH. The neural bases of complex tool use inhumans. Trends Cognit Sci. 2004;8(2):71–8.

95. Krienen FM, Buckner RL. Segregated fronto-cerebellar circuitsrevealed by intrinsic functional connectivity. Cerebr Cortex.2009;19(10):2485–97.

96. Imamizu H. Learning and switching of internal models fordexterous tool use. In: Danion F, Latash M, editors. Motorcontrol: theories, experiments, and applications. New York:Oxford Press; 2011. p. 245–266.

97. Imamizu H, Kuroda T, Miyauchi S, Yoshioka T, Kawato M.Modular organization of internal models of tools in the humancerebellum. Proc Natl Acad Sci USA. 2003;100(9):5461–6.

98. Higuchi S, Imamizu H, Kawato M. Cerebellar activity evoked bycommon tool-use execution and imagery tasks: an fMRI study.Cortex. 2007;43(3):350–8.

99. Imamizu H, Kawato M. Neural correlates of predictive andpostdictive switching mechanisms for internal models. J Neuro-sci. 2008;28(42):10751.

100. Imamizu H, Kuroda T, Yoshioka T, Kawato M. Functionalmagnetic resonance imaging examination of two modulararchitectures for switching multiple internal models. J Neurosci.2004;24(5):1173–81.

101. Dias-Ferreira E, Sousa N, Costa RM. Frontocerebellar connec-tivity: climbing through the inferior olive. Front Neurosci. 2010,3:18.

102. Depue BE, Burgess GC, Willcutt EG, Bidwell L, Ruzic L,Banich MT. Symptom-correlated brain regions in young adultswith combined-type ADHD: their organization, variability, andrelation to behavioral performance. Psychiatr Res Neuroimaging.2010;182(2):96–102.

Cerebellum

103. Durston S, van Belle J, de Zeeuw P. Differentiating frontostriataland fronto-cerebellar circuits in attention-deficit/hyperactivitydisorder. Biol Psychiatry. 2011;69(12):1178–84.

104. Svensson H, Ziemke T. Embodied representation: what are theissues. In: Bara B, Barsalou L, Buccarelli E, editors. Proceedingsof the 27th annual meeting of the Cognitive Science Society. NJ:Lawrence Erlbaum; 2005. p. 2116–2121.

105. Squire LR. Memory systems of the brain: a brief history andcurrent perspective. Neurobiol Learn Mem. 2004;82(3):171–7.

106. Poldrack RA, Clark J, Pare-Blagoev EJ, Shohamy D, MoyanoJC, Myers C, et al. Interactive memory systems in the humanbrain. Synthesis. 2001;52:297–314.

107. Doya K. What are the computations of the cerebellum, the basalganglia and the cerebral cortex? Neural Network. 1999;12(7–8):961–74.

108. Hoshi E, Tremblay L, Feger J, Carras PL, Strick PL. Thecerebellum communicates with the basal ganglia. Nat Neurosci.2005;8(11):1491–3.

109. Bostan AC, Dum RP, Strick PL. The basal ganglia communicatewith the cerebellum. Proc Natl Acad Sci USA. 2010;107(18):8452–6.

110. Rogers TD, Dickson PE, Heck DH, Goldowitz D, Mittleman G,Blaha CD. Connecting the dots of the cerebro-cerebellar role incognitive function: neuronal pathways for cerebellar modulationof dopamine release in the prefrontal cortex. Synapse. 2011;65(11):1204–12.

111. Jeannerod M. The representing brain: neural correlates of motorintention and imagery. Behav Brain Sci. 1994;17(02):187–202.

112. Jeannerod M. Neural simulation of action: a unifying mechanismfor motor cognition. NeuroImage. 2001;14(1):S103–9.

113. Wadsworth HM, Kana RK. Brain mechanisms of perceivingtools and imagining tool use acts: a functional MRI study.Neuropsychologia. 2011;49(7):1863–9.

114. Miall RC. Connecting mirror neurons and forward models.NeuroReport. 2003;14(17):2135.

115. Hotz-Boendermaker S, Hepp-Reymond MC, Curt A, Kollias SS.Movement observation activates lower limb motor networks inchronic complete paraplegia. Neurorehabil Neural Repair.2011;25(5):469.

116. Johnson-Frey SH, Newman-Norland R, Grafton ST. A distribut-ed network in the left cerebral hemisphere for planning everydaytool use actions. Cerebr Cortex. 2005;15(6):681–95.

117. Zola SM. The neurobiology of recovered memory. In: SallowaySP, Malloy PF, Cummings JL, editors. The neuropsychiatry oflimbic and subcortical disorders. Washington: American Psychi-atric Press; 1997. p. 143–54.

118. Raj V, Bell MA. Cognitive processes supporting episodicmemory formation in childhood: the role of source memory,binding, and executive functioning. Dev Rev. 2010;30(4):384–402.

119. Bachevalier J, Malkova L, Beauregard M. Multiple memorysystems: a neuropsychological and developmental perspective.In: Lyon GR, Krasnegor NA, editors. Attention, memory, andexecutive function. Baltimore: Brookes; 1996. p. 185–198.

120. Hayne H, Imuta K. Episodic memory in 3- and 4-year-oldchildren. Dev Psychobiol. 2011;53(3):317–22.

121. Poore MA, Barlow SM. Suck predicts neuromotor integrity anddevelopmental outcomes. Perspect Speech Sci Orofacial Disord.2009;19(1):44.

122. Piek JP, Dawson L, Smith LM, Gasson N. The role of early fineand gross motor development on later motor and cognitiveability. Hum Mov Sci. 2008;27(5):668–81.

123. Westendorp M, Hartman E, Houwen S, Smith J, Visscher C. Therelationship between gross motor skills and academic achieve-ment in children with learning disabilities. Research in develop-mental disabilities (in press) corrected proof; 2011.

124. Largo RH, Fischer JE, Rousson V. Neuromotor developmentfrom kindergarten age to adolescence: developmental course andvariability. Swiss Med Wkly. 2003;133(13/14):193–9.

125. Gibson KR. Evolution of human intelligence: the roles of brainsize and mental construction. Brain Behav Evol. 2002;59(1–2):10–20.

126. Blumenfeld H. Neuroanatomy through clinical cases. SinauerAssociates; 2002.

127. Von Hofsten C. Action in development. Dev Sci. 2007;10(1):54–60.

128. Von Hofsten C. Action, the foundation for cognitive development.Scand J Psychol. 2009;50(6):617–23.

129. Lockman JJ. A perception-action perspective on tool usedevelopment. Child Dev. 2000;71(1):137–44.

130. Tomasello M, Carpenter M, Liszkowski U. A new look at infantpointing. Child Dev. 2007;78(3):705–22.

131. Want SC, Harris PL. Learning from other people's mistakes:causal understanding in learning to use a tool. Child Dev.2001;72(2):431–43.

132. Piek JP. Infant motor development. Human Kinetics Publishers;2006.

133. Von Hofsten C. An action perspective on motor development.Trends Cogn Sci. 2004;8(6):266–72.

134. Adolph KE, Berger SA. Motor development. In: Damon W,Lerner R, series editors; and Kuhn D, Siegler S, volume editors.Handbook of child psychology: Vol 2: Cognition, perception,and language (6th ed.) New York: Wiley; 2006. p. 161–213.

135. Koziol LF, Budding DE. Subcortical structures and cognition:implications for neuropsychological assessment. New York:Springer; 2009.

136. Cotterill RM. Cooperation of the basal ganglia, cerebellum,sensory cerebrum and hippocampus: possible implications forcognition, consciousness, intelligence and creativity. Prog Neu-robiol. 2001;64(1):1–33.

137. Hazy TE, Frank MJ, O'Reilly RC. Banishing the homunculus:making working memory work. Neuroscience. 2006;139(1):105–18.

138. Houk JC, Wise SP. Distributed modular architectures linkingbasal ganglia, cerebellum, and cerebral cortex: their role inplanning and controlling action. Cereb Cortex. 1995;5(2):95–110.

139. Imamizu H, Kawato M. Cerebellar internal models: implicationsfor the dexterous use of tools. Cerebellum. 2010;22:1–11.

140. Milner TE, Franklin DW, Imamizu H, Kawato M. Centralcontrol of grasp: manipulation of objects with complex andsimple dynamics. NeuroImage. 2007;36(2):388–95.

141. Schultz J, Imamizu H, Kawato M, Frith CD. Activation of thehuman superior temporal gyrus during observation of goalattribution by intentional objects. J Cogn Neurosci. 2004;16(10):1695–705.

142. Panksepp J. Affective neuroscience: the foundations of humanand animal emotions. USA: Oxford University Press; 2004.

143. De Quiros JB, Schrager, OL. Neuropsychological fundamentalsin learning disabilities. Novato, CA: Academic Therapy; 1979.

144. Rorke LB, Riggs HE. Myelination of the brain in the newborn.Philadelphia: Lippincott; 1969.

145. Power JD, Fair DA, Schlaggar BL, Petersen SE. The develop-ment of human functional brain networks. Neuron. 2010;67(5):735–48.

146. Yakovlev PI, Lecours AR. The myelogenetic cycles of regionalmaturationh of the brain. In: Minkowski A, editor. Regionaldevelopment of the brain in early life. Boston: BlackwellScientific; 1967. p. 3–70.

147. Altman J, Bayer SA. Development of the cerebellar system: inrelation to its evolution, structure, and functions. Boca Raton:CRC; 1997.

Cerebellum

148. Giedd JN, Rapoport JL. Structural MRI of pediatric braindevelopment: what have we learned and where are we going?Neuron. 2010;67(5):728–34.

149. Barber AD, Srinivasan P, Joel SE, Caffo BS, Pekar JJ,Mostofsky SH. Motor “Dexterity”?: evidence that left hemi-sphere lateralization of motor circuit connectivity is associatedwith better motor performance in children. Cerebral Cortex 2011(in press)

150. Petrini JR, Dias T, McCormick MC, Massolo ML, Green NS,Escobar GJ. Increased risk of adverse neurological developmentfor late preterm infants. J Pediatr. 2009;154(2):169–76.

151. Bhutta AT, Cleves MA, Casey PH, Cradock MM, Anand KJ.Cognitive and behavioral outcomes of school-aged children whowere born preterm: a meta-analysis. JAMA. 2002;288(6):728–37.

152. Riva D, Vago C, Usilla A, Treccani C, Pantaleoni C, D'Arrigo S,et al. The role of the cerebellum in higher cognitive and socialfunctions in congenital and acquired diseases of developmentalage. In: Riva D, Njiokiktjien C, editors. Brain lesion localizationand developmental functions. Montrouge: John Libbey Eurotext;2010. p. 133–44.

153. Limperopoulos C, Soul JS, Haidar H, Huppi PS, Bassan H,Warfield SK, et al. Impaired trophic interactions between thecerebellum and the cerebrum among preterm infants. Pediatrics.2005;116(4):844–50.

154. Messerschmidt A, Brugger PC, Boltshauser E, Zoder G,Sterniste W, Birnbacher R, et al. Disruption of cerebellardevelopment: potential complication of extreme prematurity.AJNR Am J Neuroradiol. 2005;26(7):1659–67.

155. Allin M, Matsumoto H, Santhouse AM, Nosarti C, AlAsadyMH, Stewart AL, et al. Cognitive and motor function and thesize of the cerebellum in adolescents born very pre-term. Brain.2001;124(Pt 1):60–6.

156. Allin MP, Salaria S, Nosarti C, Wyatt J, Rifkin L, Murray RM.Vermis and lateral lobes of the cerebellum in adolescents bornvery preterm. NeuroReport. 2005;16(16):1821–4.

157. Herbert JS, Eckerman CO, Goldstein RF, Stanton ME. Contrastsin infant classical eyeblink conditioning as a function ofpremature birth. Infancy. 2004;5(3):367–83.

158. Kessenich M. Developmental outcomes of premature, low birthweight, and medically fragile infants. Newborn and Infant NursRev. 2003;3(3):80–7.

159. Volpe JJ. Cerebellum of the premature infant: rapidly develop-ing, vulnerable, clinically important. J Child Neurol. 2009;24(9):1085–104.

160. Tanskanen P, Valkama M, Haapea M, Barnes A, Ridler K,Miettunen J, et al. Is prematurity associated with adult cognitiveoutcome and brain structure? Pediatr Neurol. 2011;44(1):12–20.

161. Haldipur P, Bharti U, Alberti C, Sarkar C, Gulati G, Iyengar S, etal. Preterm delivery disrupts the developmental program of thecerebellum. PLoS One. 2011;6(8):e23449.

162. Connolly KJ, Dalgleish M. Individual patterns of tool use byinfants. In: Kalverboer AF, Hopkins B, Geuze R, editors. Motordevelopment in early and later childhood: longitudinal

approaches. Cambridge: Cambridge University Press; 1993;p. 174–204.

163. Voogd J, Schraa-Tam CKL, van der Geest JN, De Zeeuw CI.Visuomotor cerebellum in human and nonhuman primates.Cerebellum 2010;1–19.

164. Blakemore SJ, Sirigu A. Action prediction in the cerebellum andin the parietal lobe. Exp Brain Res. 2003;153(2):239–45.

165. Doron KW, Funk CM, Glickstein M. Fronto-cerebellar circuitsand eye movement control: a diffusion imaging tractographystudy of human cortico-pontine projections. Brain Res.2010;1307:63–71.

166. Diamond A. Close interrelation of motor development andcognitive development and of the cerebellum and prefrontalcortex. Child Dev. 2000;71(1):44–56.

167. Li S, Ostwald D, Giese M, Kourtzi Z. Flexible coding forcategorical decisions in the human brain. J Neurosci. 2007;27(45):12321.

168. Banich MT, Compton RJ. Cognitive neuroscience. 3rd ed.Belmont: Wadsworth; 2011.

169. Seger CA. The involvement of corticostriatal loops in learningacross tasks, species, and methodologies. In: Groenewegen H,Berendse H, editors. The Basal Ganglia IX: Proceedings of the9th Triennial Meeting of the International Basal Ganglia Society.New York: Springer; 2009.

170. Seger CA, Miller EK. Category learning in the brain. Annu RevNeurosci. 2010;Jul 21;33(1).

171. Schmahmann JD. The cerebellum and cognition. San Diego:Academic Press; 1997.

172. Higuchi S, Chaminade T, Imamizu H, Kawato M. Shared neuralcorrelates for language and tool use in Broca's area. Neuro-Report. 2009;20(15):1376–81.

173. Stout D, Toth N, Schick K, Chaminade T. Neural correlates ofEarly Stone Age toolmaking: technology, language and cognitionin human evolution. Philos Trans R Soc Lond B Biol Sci.2008;363(1499):1939–49.

174. Vandervert LR, Koziol LF. How language came to be. Posterpresented at the Second International Congress of the Society forResearch on the Cerebellum. Chicago, IL. 2009.

175. Ullman MT, Pierpont EI. Specific language impairment is notspecific to language: the procedural deficit hypothesis. Cortex.2005;41(3):399–433.

176. Ullman MT. Contributions of memory circuits to language:the declarative/procedural model. Cognition. 2004;92(1–2):231–70.

177. Glenberg AM, Gallese V. Action-based language: a theory oflanguage acquisition, comprehension, and production. Cortex.2011. doi:10.1016/j.cortex.2011.04.010.

178. Gallese V, Rochat M, Cossu G, Sinigaglia C. Motor cognitionand its role in the phylogeny and ontogeny of action under-standing. Dev Psychol. 2009;45(1):103.

179. Mostofsky SH, Ewen JB. Altered connectivity and action modelformation in autism is autism. Neuroscientist. 2011. doi:10.1177/1073858410392381.

180. Schmahmann JD, Sherman JC. Cerebellar cognitive affectivesyndrome. Int Rev Neurobiol. 1997;41:433–40.

Cerebellum